The goal of the proposed research is to advance the study of vulnerability to substance abuse by bridging the gap between genes and behavior via sophisticated analyses of copy number variation (CNV) and brain activation. Three substances, nicotine, alcohol, and marijuana, will be investigated using multimodality data from multiple studies. CNV covers large amounts of nucleotide sequence variation, through deletion and insertion of DNA segments ranging from 1 kilobase to several megabases. Functional magnetic resonance imaging (fMRI) provides spatially localized, neurobiological information. Behavior and neuropsychological assessments provide information about symptoms and other variables. Integrating information at different phases of progressive development of a substance disorder, a multilevel vulnerability model of substance abuse will be established to identify pathways from genes to biology and finally to behavior. The common and distinct pathways among different types of substance abuse will also be studied. The increased knowledge of how the brain interacts with genes and behavior will significantly advance the diagnosis of a substance abuse disorder and assist prevention plans via monitoring and intervening as early as possible. In addition, the knowledge of common pathways among substance abuse will improve the overall effectiveness of diagnosis, prevention and treatment. The proposed research is divided into two phases. The R21 phase will focus on development of a multivariate genomic association analysis method, termed parallel independent component analysis (ICA). The relationship between genomic CNV array data and fMRI brain activation will be accessed via the parallel ICA through simulation and real application, compared with the often-used univariate correlation test. The parallel ICA method will be implemented to data from both drinkers and smokers, and a test- replication procedure will verify the identified CNV-fMRI associations. The goal of the R33 phase is to conduct a study of vulnerability to substance abuse, including a multilevel vulnerability model from genetics to brain to behavior and common pathways among different types of substance abuse. The first step in the R33 phase is to extract all three-way relations between CNV, fMRI brain activation, and behavioral assessments. Then, the relations will be factored into a multilevel model, where consecutive CNV-brain-behavior connections and segmental relations are identified. After independently conducting the multilevel vulnerability study on alcohol, nicotine, and marijuana abuse, the common pathways among them will be extracted.